External/cglm
1
0
Fork 0
mirror of https://github.com/recp/cglm.git synced 2025-12-25 12:55:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

arm, neon: neon/fma support for glm_mul()

Browse Source
This commit is contained in:
Recep Aslantas
2021-04-25 03:45:00 +03:00
parent 0ab50f7208
commit 2903813765
3 changed files with 58 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
include/cglm/affine-mat.h
View File

@@ -26,6 +26,10 @@
# include "simd/avx/affine.h"
#endif
#ifdef CGLM_NEON_FP
# include "simd/neon/affine.h"
#endif
/*!
* @brief this is similar to glm_mat4_mul but specialized to affine transform
*
@@ -49,6 +53,8 @@ glm_mul(mat4 m1, mat4 m2, mat4 dest) {
glm_mul_avx(m1, m2, dest);
#elif defined( __SSE__ ) || defined( __SSE2__ )
glm_mul_sse2(m1, m2, dest);
#elif defined(CGLM_NEON_FP)
glm_mul_neon(m1, m2, dest);
#else
float a00 = m1[0][0], a01 = m1[0][1], a02 = m1[0][2], a03 = m1[0][3],
a10 = m1[1][0], a11 = m1[1][1], a12 = m1[1][2], a13 = m1[1][3],

51
include/cglm/simd/neon/affine.h Normal file
View File

@@ -0,0 +1,51 @@
/*
* Copyright (c), Recep Aslantas.
*
* MIT License (MIT), http://opensource.org/licenses/MIT
* Full license can be found in the LICENSE file
*/
#ifndef cglm_affine_neon_h
#define cglm_affine_neon_h
#if defined(__ARM_NEON_FP)
#include "../../common.h"
#include "../intrin.h"
CGLM_INLINE
void
glm_mul_neon(mat4 m1, mat4 m2, mat4 dest) {
/* D = R * L (Column-Major) */
glmm_128 l0, l1, l2, l3, r0, r1, r2, r3, v0, v1, v2, v3;
l0 = glmm_load(m1[0]); r0 = glmm_load(m2[0]);
l1 = glmm_load(m1[1]); r1 = glmm_load(m2[1]);
l2 = glmm_load(m1[2]); r2 = glmm_load(m2[2]);
l3 = glmm_load(m1[3]); r3 = glmm_load(m2[3]);
v0 = vmulq_f32(glmm_splat_x(r0), l0);
v1 = vmulq_f32(glmm_splat_x(r1), l0);
v2 = vmulq_f32(glmm_splat_x(r2), l0);
v3 = vmulq_f32(glmm_splat_x(r3), l0);
v0 = glmm_fmadd(glmm_splat_y(r0), l1, v0);
v1 = glmm_fmadd(glmm_splat_y(r1), l1, v1);
v2 = glmm_fmadd(glmm_splat_y(r2), l1, v2);
v3 = glmm_fmadd(glmm_splat_y(r3), l1, v3);
v0 = glmm_fmadd(glmm_splat_z(r0), l2, v0);
v1 = glmm_fmadd(glmm_splat_z(r1), l2, v1);
v2 = glmm_fmadd(glmm_splat_z(r2), l2, v2);
v3 = glmm_fmadd(glmm_splat_z(r3), l2, v3);
v3 = glmm_fmadd(glmm_splat_w(r3), l3, v3);
glmm_store(dest[0], v0);
glmm_store(dest[1], v1);
glmm_store(dest[2], v2);
glmm_store(dest[3], v3);
}
#endif
#endif /* cglm_affine_neon_h */

2
include/cglm/simd/neon/mat4.h
View File

@@ -41,7 +41,7 @@ glm_mat4_transp_neon(mat4 m, mat4 dest) {
CGLM_INLINE
void
glm_mat4_mul_neon(mat4 m1, mat4 m2, mat4 dest) {
/* D = R * L (Column-Major) */
/* D = R * L (Column-Major) */
glmm_128 l0, l1, l2, l3, r0, r1, r2, r3, v0, v1, v2, v3;